Blood Pressure Estimation Based on Photoplethysmography for Personalized Healthcare

Cuff-less BP measurement methods suitable for IoT applications have been of specific interest for researchers of late. However, most of the methods are based on use of electrocardiograph (ECG) signal along with PPG signal or using extensive learning and artificial intelligence (AI). In this paper two features of the PPG signal viz. peak to peak amplitude $(v_{PP})$ and foot to foot delay $(D)$ have been used to measure first the diastolic blood pressure $(P_{D})$ and then the systolic blood pressure $(P_{S})$ . A novel expression is derived from Beer Lambert's law to relate $P_{D}$ with $v_{PP}$ . A two-pulse-synthesis (TPS) model is used to decompose a PPG pulse using two Rayleigh functions and foot to foot delay is extracted. $P_{S}$ has been obtained using $D$ . The method has been tested on 31 volunteers and 150 diseased subjects from MIMIC III waveform database. Mean absolute error of $P_{S}$ and $P_{D}$ for MIMIC III database are 3.63 mmHg and 2.28 mmHg respectively which are very much comparable to other popular calibration based methods reported in the literature. This lightweight method may be suitable for personalized healthcare.